nx_gonogo.nx

nx_gonogo.nx -- THE dmkt CAPSTONE (F317 / milestone M9): the GO / NO-GO decision gate. The lane's charter claims a CLOSED idea-to-market loop. Without this organ the loop was open: nx_product_score could score CAPABILITY COVERAGE but had no economics, and nx_sourcing_path could price a sourcing mix but had no verdict. This joins them. ★★THE DESIGN HINGES ON A THIRD VERDICT. A gate that can only answer GO or NO-GO **will fabricate**: asked to decide with a missing input it must invent one, and the invented one is always the one nobody can check. So the verdict space is GO | NO-GO | BLOCKED-ON-EVIDENCE, and the blocked state NAMES the evidence it lacks. This is the operator's no-fake-numbers law expressed as control flow rather than as a comment. ★★THE ASYMMETRY IS THE WHOLE VALUE. Missing evidence blocks a GO but must NEVER block a NO-GO. A product whose priced mix loses money is rejectable TODAY, with no demand study, no market research and no propensity model -- you do not need to know how many people want a thing to know you cannot afford to make it. Checking hard failures FIRST is what stops this gate degenerating into a permanently-blocked oracle that answers nothing. Cheap disqualification is a feature. DATA, NOT CODE (rule 11): every threshold is a plane row. There is not one hardcoded bar here. `gonogothresh-` : <id> TAB <label> TAB <value> margin_floor_permil | domestic_floor_permil | risk_ceiling | coverage_floor_permil `demand-` : ANY row = demand evidence exists. ZERO rows = the demand gate is BLIND. Set a threshold to 0 to disable that gate deliberately -- an explicit, auditable waiver. Composes nx_sourcing_path (which composes nx_landed_cost) -- one duty implementation, one mix pricer, no copies. Target-imports-target is proven and gated by nx_lcimport_test. VERBS: nx_gonogo verdict <bomprefix> <units> <unitprice_c> <ocean01> <pref01> <incprefix> <threshprefix> <demandprefix> [coverage_permil|-1] nx_gonogo selftest license_tier: ORIGINAL No hw writes (Rule 26). expect_exit: 0